27
Execution Environments for Distributed Computing Intelligent Placement of Datacenters for Internet Services EEDC 3 4 3 3 0 Master in Computer Architecture, Networks and Systems - CANS Homework number: 6 Members: Roger Rafanell [email protected]

EEDC Intelligent Placement of Datacenters

Embed Size (px)

Citation preview

Page 1: EEDC Intelligent Placement of Datacenters

Execution Environments for Distributed Computing

Intelligent Placement of Datacenters for Internet

Services

EEDC

343

30

Master in Computer Architecture, Networks and Systems - CANS

Homework number: 6

Members: Roger Rafanell [email protected]

Page 2: EEDC Intelligent Placement of Datacenters

2

Outline

Introduction Motivation Placing Datacenters Evaluation Conclusion

2

Page 3: EEDC Intelligent Placement of Datacenters

3

Introduction: Datacenter construction costs

Each datacenter costs >$100M to construct– The smaller datacenters are rated at ~25MW

Examples:– Microsoft DCs in Virginia & Chicago: $500M each

3

Page 4: EEDC Intelligent Placement of Datacenters

4

Outline

Introduction Motivation Placing Datacenters Evaluation Conclusion

4

Page 5: EEDC Intelligent Placement of Datacenters

5

Motivation

Internet services require thousands of servers

Use multiple “mirror” datacenters– High availability and fault tolerance

– Low response time

Spend millions building and operating datacenters

Consume enormous amounts of brown energy!!

Page 6: EEDC Intelligent Placement of Datacenters

6

Outline

Introduction Motivation Placing Datacenters Evaluation Conclusion

6

Page 7: EEDC Intelligent Placement of Datacenters

7

Intelligent placement of datacenters

Goal: Manage the monetary and environmental costs

Define framework Model costs and datacenter characteristics Create solution approaches Collect cost and location-related data Create placement tool

7

Page 8: EEDC Intelligent Placement of Datacenters

8

Selecting datacenter locations

Model datacenter placement– Network latencies

– Availability

8

Page 9: EEDC Intelligent Placement of Datacenters

9

Selecting datacenter locations

Model datacenter placement– Network latencies

– Availability

CAPEX costs– Distance to electricity and networking infrastructure

– Land and construction (maximum PUE)

– Power delivery, cooling, backup equipment

– Servers and networking equipment

9

Page 10: EEDC Intelligent Placement of Datacenters

10

Selecting datacenter locations

Model datacenter placement– Network latencies

– Availability

CAPEX costs– Distance to electricity and networking infrastructure

– Land and construction (maximum PUE)

– Power delivery, cooling, backup equipment

– Servers and networking equipment

OPEX costs– Maintenance and administration

– Electricity and water prices (average PUE)

10

Page 11: EEDC Intelligent Placement of Datacenters

11

Selecting datacenter locations

Model datacenter placement– Network latencies

– Availability

CAPEX costs– Distance to electricity and networking infrastructure

– Land and construction (maximum PUE)

– Power delivery, cooling, backup equipment

– Servers and networking equipment

OPEX costs– Maintenance and administration

– Electricity and water prices (average PUE)

Incentives (taxes)

11

Page 12: EEDC Intelligent Placement of Datacenters

12

Selecting datacenter locations

Model datacenter placement– Network latencies

– Availability

CAPEX costs– Distance to electricity and networking infrastructure

– Land and construction (maximum PUE)

– Power delivery, cooling, backup equipment

– Servers and networking equipment

OPEX costs– Maintenance and administration

– Electricity and water prices (average PUE)

Incentives (taxes)

12

Page 13: EEDC Intelligent Placement of Datacenters

13

The problem formulation

Goal– Minimize CAPEX and OPEX

Constraints– Response times < MAX LATENCY for all users– Min consistency delay between 2 DCs < MAX DELAY– Min system availability > MIN AVAILABILITY

Output– Number of servers at each location– Minimum cost

13

Page 14: EEDC Intelligent Placement of Datacenters

14

Solving the (non-linear) problem

Linear Programming– Does not support non-linear costs

Brute force– Too slow

Simple heuristics– May not produce accurate results efficiently

14

Page 15: EEDC Intelligent Placement of Datacenters

15

Our approach for solving the problem

Evaluate each potential solution– Quickly via Linear Programming (LP)

Consider neighboring configurations– Simulated annealing (SA)

Cost optimization process– Combine SA and LP

15

Current solution Near neighbor

LP

SA

LP

Page 16: EEDC Intelligent Placement of Datacenters

16

Our approach for solving the problem

16

LP

SA

LP

LP

SA

LP

SA

$13.8M/month

$9.2M/month $10.7M/month

$10.3M/month

Page 17: EEDC Intelligent Placement of Datacenters

17

Summary of our approach

1) Generate a grid of tentative locations

2) Collect data about each location

3) Define datacenter characteristics

4) Instantiate optimization problem

5) Solve optimization problem

17

Page 18: EEDC Intelligent Placement of Datacenters

18

Outline

Introduction Motivation Placing Datacenters Evaluation Conclusion

18

Page 19: EEDC Intelligent Placement of Datacenters

19

Comparing locations for 60k-server DC

0100020003000400050006000700080009000

Austin Bismarck Los Angeles

New York Orlando Seattle St. LouisCost

(tho

usan

d do

llars

per

mon

th)

Servers Land Building Connection Energy Water Staff Networking

19

Page 20: EEDC Intelligent Placement of Datacenters

20

Interesting questions

How much does…… lower latency cost?

… higher availability cost?

… faster consistency cost?

… a green DC network cost?

… a chiller-less DC network cost?

20

Page 21: EEDC Intelligent Placement of Datacenters

21

Cost of 60k-server green DC network

21

Green DC network costs $100k/month more, except when latency <70ms

Page 22: EEDC Intelligent Placement of Datacenters

22

Cost of a 60k-server chiller-less DC network

0

2

4

6

8

10

12

14

30 50 70 90 110

Cost

(in

mill

ion

dolla

rs)

Maximum latency (milliseconds)

Chiller-less

Traditional

22

Chiller-less DC network is cheaper but it cannot achieve low latencies

Page 23: EEDC Intelligent Placement of Datacenters

23

Conclusions

First scientific work on smart datacenter placement– Proposed framework and optimization problem

– Proposed solution approach

– Characterized many locations across the US

– Built a tool to automate the process

– Answered many interesting questions

Results show that smart placement can save millions Work enables smaller companies to reap the benefits

23

Page 24: EEDC Intelligent Placement of Datacenters

24

Future work

To extend with data from Europe

Include tax incentives

Test the tool with data from real services

24

Page 25: EEDC Intelligent Placement of Datacenters

25

Questions

Page 26: EEDC Intelligent Placement of Datacenters

26

Specially thanks to the real authors of the work …

Íñigo Goiri, Kien Le, Jordi Guitart,Jordi Torres, and Ricardo Bianchini

26

Page 27: EEDC Intelligent Placement of Datacenters

27

Energy costs and carbon emissions

Company #ServersEnergy/year

(MWh)Energy

cost/yearCO2/year

(Metric tons)

eBay 16K 0.6 x 105 $3.7M 0.4 x 105

Akamai 40K 1.7 x 105 $10M 1.0 x 105

Rackspace 50K 2 x 105 $12M 1.2 x 105

Microsoft >200K >6 x 105 >$36M >3.6 x 105

Google >500K >6.3 x 105 >$38M >3.8 x 105

Sources: [Qureshi’09], EPA

27